OPEN Citation: Cell Discovery (2017) 3, 17034; doi:10.1038/celldisc.2017.34 ARTICLE www.nature.com/celldisc

Systematic analysis of human telomeric dysfunction using inducible telosome/shelterin CRISPR/Cas9 knockout cells

Hyeung Kim1, Feng Li2, Quanyuan He1, Tingting Deng2, Jun Xu3, Feng Jin4, Cristian Coarfa4, Nagireddy Putluri4, Dan Liu1,3, Zhou Songyang1,2,*

1Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, TX, USA; 2Key Laboratory of Gene Engineering of the Ministry of Education and State Key Laboratory for Biocontrol, School of Life Sciences, Sun Yat-sen University, Guangzhou, China; 3Cell-Based Assay Screening Service Core, Baylor College of Med- icine, Houston, TX, USA; 4Department of Molecular and Cellular Biology and Advanced Technology Core, Baylor College of Medicine, Houston, TX, USA

CRISPR/Cas9 technology enables efficient loss-of-function analysis of human genes using somatic cells. Studies of essential genes, however, require conditional knockout (KO) cells. Here, we describe the generation of inducible CRISPR KO human cell lines for the subunits of the telosome/shelterin complex, TRF1, TRF2, RAP1, TIN2, TPP1 and POT1, which directly interact with or can bind to telomeres through association with other subunits. Homozygous inactivation of several subunits is lethal in mice, and most loss-of-function studies of human regulators have relied on RNA interference-mediated gene knockdown, which suffers its own limitations. Our inducible CRISPR approach has allowed us to more expediently obtain large numbers of KO cells in which essential telomere regulators have been inac- tivated for biochemical and molecular studies. Our systematic analysis revealed functional differences between human and mouse telomeric proteins in DNA damage responses, telomere length and metabolic control, providing new insights into how human telomeres are maintained. Keywords: CRISPR/Cas9; inducible knockout; metabolism; POT1 isoform; telomere; telosome/shelterin Cell Discovery (2017) 3, 17034; doi:10.1038/celldisc.2017.34; published online 26 September 2017

Introduction instance, TRF1 and TRF2 bind directly to the telomere duplex through their myb domains [9–13], whereas In the past 20 years, we have gained tremendous POT1 binds 3’ single-stranded (ss) telomeric overhangs insight into how the ends of mammalian chromosomes [14, 15]. RAP1 is recruited by TRF2, but apparently does or telomeres are maintained and regulated. Together not directly interact with any of the other subunits [16]. with the , which consists of the reverse tran- TIN2 can interact with both TRF1 and TRF2 [6, 17–21]. scriptase TERT and RNA template TR/TERC, a mul- It also binds TPP1 and helps bring to telomeres the titude of telomere-binding proteins participate in TPP1-POT1 heterodimer that is essential for regulating – telomere maintenance [1 5]. In particular, six core telo- telomerase access to telomeres [21–30]. The core telo- meric proteins, TRF1, TRF2, RAP1, TPP1, TIN2 and mere proteins often act as interaction hubs to recruit POT1, dynamically assemble on telomeres as a large factors of diverse pathways to telomeres and ensure complex called telosome or shelterin and are essential in crosstalk between telomere maintenance pathways and telomere length regulation and end protection in mam- other cellular processes [8, 19, 31, 32]. In fact, several key – mals [6 8]. Extensive research has revealed the interac- telomere regulators have been shown to regulate meta- tions and functions of telosome components. For bolism, providing direct evidence of the close ties between telomere regulation and metabolic control. For *Correspondence: Zhou Songyang example, the human telomerase reverse transcriptase has Tel: +713 798 5220; Fax: +713 796 9438 E-mail: [email protected] been found to localize to the mitochondria and reduce Received 15 March 2017; accepted 27 July 2017 intracellular oxidative stress [33–36]. Our lab has found Studying telosome subunits using inducible knockout cell lines 2 that TIN2 can also localize to the mitochondria and of their human orthologs, we first turned to RNAi KD regulate oxidative phosphorylation [37]. in human cells through stable expression of short Numerous studies have demonstrated that dysfunc- hairpin RNAs (Supplementary Figure S1A). Even with tional telomeres can lead to telomere length defects, effective KD (480%) of TRF2, for example, we could deprotected telomeres, genomic instability and diseases only observe minor DNA damage responses (DDRs) [1, 4, 32, 38]. Much of our knowledge regarding the at telomeres (data not shown), rarely more severe molecular and functional significance of mammalian telo- phenotypes such as chromosome end-to-end fusions meric proteins comes from studies using mouse knockout found in Trf2 KO MEF cells [50], suggesting that (KO) mouse embryonic fibroblast (MEF) cells, as genes residual TRF2 proteins in the KD cells may have been are more readily targeted in mouse embryonic stem cells. sufficient to prevent severe and sustained telomere However, notable differences exist in telomere regulation DNA damages. We next attempted straight KO of between mouse and human. For instance, human telo- these genes by CRISPR/Cas9, but failed to isolate any meres are considerably shorter than those of laboratory clones of TRF2, TIN2 or POT1 KO cells. Given such mice and human has one POT1 gene, whereas mouse has findings, we decided that human cells conditionally two (Pot1a and Pot1b). Such disparities underscore the knocked out for telosome subunits would be more need for loss-of-function human cellular models. Majority desirable. of the loss-of-function studies in human cells have relied on Traditionally, conditional KO alleles are generated RNA interference (RNAi)-mediated inhibition of endo- by inserting into some particular locus recombination genous genes. The limitations of RNAi knockdown (KD) sequences such as loxp sites, which can mediate the and the fact that several key telomere regulators including deletion of intervening sequences upon the expression TRF2 and TIN2 are essential genes have complicated data of recombinases such as Cre [51–53]. The insertion of analysis and interpretation. Complete inactivation of these such exogenous sequences may alter gene regulation telomere regulatory genes in cells may cause cell death, and the entire process is often time consuming. To precluding further detailed biochemical and molecular generate conditional telosome subunit KO cells, we studies, especially experiments that require extended cul- modified a lentivirus-based inducible CRISPR/Cas9 turing and/or large numbers of cells. KO system [54], and constructed separate vectors for The advent of the CRISPR/Cas9 genome-editing tech- inducible Cas9 and constitutive single guide RNA nology has afforded investigators unprecedented opportu- (sgRNA) expression (Supplementary Figure S1B). nities to more efficiently and specifically target genes in Hela cells were transduced with lentiviruses encoding human cells and to explore the consequences of their inducible Cas9 and a clone in which Cas9 expression inactivation [39–47]. In this study, we took advantage of the could be reproducibly activated with doxycycline in a highly flexible and adaptable CRISPR/Cas9 system and dose-dependent manner was selected (Supplementary generated human inducible KO cell lines for each of the Figure S1C). telosome components. This panel of cells has allowed us to The double-strand breaks resulting from Cas9 clea- survey the functional significance of each telomeric protein vage trigger the non-homologous end joining DNA and probe the impact of individual subunit inhibition on repair pathway in the absence of a donor template telomere regulation as well as metabolic control. With this [55–57]. Non-homologous end joining-mediated DNA systematic analysis of the function of human telomere repair may generate small insertions and/or deletions proteins using inducible KO cell lines, we are able to better (indels) at the target site, and compromise gene func- delineate the differences between mouse and human telo- tion if cleavage occurs within protein coding sequences. mere biology. In addition, our panel of inducible KO cell Repair of a single Cas9 cleavage site has a 1/3 chance of lines should prove invaluable to investigators seeking to in-frame ligation of the coding sequences, which may further explore the consequences of telomere dysfunction not completely disrupt gene function. We reasoned that and to study how diverse cellular functions may be dis- simultaneous targeting with two sgRNAs should rupted upon telomere dysregulation. improve the odds of larger deletions and more com- plete inhibition of endogenous genes. To test this Results strategy, the inducible Cas9 cells were infected with two viruses encoding two separate TIN2-specific sgRNAs Using CRISPR/Cas9 to generate inducible KO human either singly or together, selected with appropriate cell lines antibiotics, and then cultured in doxycycline- Trf1, Trf2 and Tin2 have been reported to be containing media to induce Cas9 expression essential genes in mouse [48–50]. To determine the roles (Figure 1a). At different time points following

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 3 cell lines showed similar growth patterns in the absence of doxycycline, differences in growth rates became apparent between doxycycline-induced TIN2 KO cells after 4-day treatment. Growth of the single sgRNA TIN2 KO cells was hampered initially, but appeared to recover with continued culturing, likely due to the presence of cells with incomplete TIN2 inhibition. Indeed, the severity of proliferation defects correlated with the degree of TIN2 ablation, with the dual sgRNA TIN2 KO cells being more severely affected than the single sgRNA KO cells (Figure 1c). When we ectopi- cally expressed sgRNA-resistant TIN2 in the dual sgRNA KO cells, growth and proliferation were restored, indicating that TIN2 was critical for cell growth and that dual sgRNAs more completely knocked out TIN2. Although the inducible TIN2 KO cells were poly- clonal, independent inductions of Cas9 led to highly reproducible results, indicating that the inducible strategy reliably produces populations of cells with comparable genotypes and phenotypes. When we sampled the TIN2 alleles from the induced dual sgRNA TIN2 KO cells by TOPO cloning and Sanger sequencing of the sgRNA target region (Supplementary Figure S2A), we found most (480%) Figure 1 The dual sgRNA strategy enables more complete to contain deletions because of simultaneous Cas9 inactivation of endogenous genes. (a) Schematic representation cleavage at both sgRNA target sites, and the remaining of the TIN2 locus. Shaded boxes indicate exons and sites targeted by the sgRNAs (g1 and g2) are marked. The distance alleles containing indels at both target sites without between the two predicted cleavage sites (red asterisks) is deleting the intervening sequences. Importantly, all of 147 bps. Red arrows indicate positions of PCR primers for the alleles are predicted to have impaired TIN2 func- genomic sequence verifications. (b) Cas9-inducible Hela cells tion, corroborating that dual sgRNA design helped stably expressing single (TIN2_g1 or TIN2_g2) or dual sgRNAs ensure complete inactivation of endogenous genes. − (TIN2_DG) were induced with 1 μgml 1 doxycycline for 3, 6 or Using the dual sgRNA system, we generated indu- 9 days, and then collected for immunoblotting using the indicated cible KO cell lines for all six core telomeric proteins antibodies. TIN2_DG cells that also stably expressed sgRNA- (Supplementary Figure S2B). Again, we compared resistant Flag-tagged TIN2 (Rescue) were also included. The fi anti-SMC1 antibody served as a loading control. The expected single vs dual sgRNA KO ef ciencies. Although some sizes for endogenous (Endo) and exogenous (Exo) TIN2 are of the single sgRNAs knocked out endogenous gene indicated by arrows. (c) Cells from b were collected at the expression quite effectively, using dual sgRNAs to indicated time points following addition of doxycycline (Dox). simultaneously target a single locus consistently proved The number of live cells was determined by Trypan blue exclu- more efficient (Supplementary Figure S2C). Again, sion. At least three experiments with independent doxycycline longer doxycycline induction led to more effective and inductions were performed and the results were combined and sustained inactivation of endogenous genes plotted as indicated. Error bars indicate s.e. (Supplementary Figure S2C). In the following experi- ments, all the cell lines were induced for 6 days with doxycycline treatment, cells were collected for analysis doxycycline before further analysis and/or treatment of TIN2 protein expression (Figure 1b). As we pre- unless otherwise specified. dicted, targeting with two sgRNAs appeared to KO more efficiently than using a single Profiling the contribution of each subunit to the sgRNA. Furthermore, lengthier doxycycline treatment telomeric assembly of the telosome complex was able to improve KO efficiency (Figure 1b). The six KO cell lines afforded us the first opportu- Notably, the TIN2 KO cells exhibited proliferative nity to systematically investigate in detail how con- defects during culturing (Figure 1c). Although all of the stitutive deletion of one subunit may affect the

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 4

Figure 2 Removal of individual subunits impacts the organization of the telosome. (a) For each subunit of the telosome, we generated inducible Cas9 cells stably expressing two sgRNAs targeting the gene. Cells were treated with doxycycline for up to 6 days before being analyzed by western blotting using the indicated antibodies. Anti-SMC1 and -actin antibodies served as loading controls. Molecular weight makers (kDa) are indicated on the right. (b) Cells from a (after 6 days of induction) were immunoprecipitated using appropriate antibodies to bring down associated telomeric DNA for dot blotting and hybridization with a

telomere repeat probe (TTAGGG)3. Signals for each cell line were quantified and normalized against input. At least three experiments with independent doxycycline inductions were performed for each cell line, and the results were combined and plotted as indicated. Error bars indicate s.e. P-values were obtained using the Student’s t-test. *Po0.05, **Po0.01.

telomeric targeting and assembly of the telosome other telosome subunits, particularly POT1, under- complex. Each cell line was induced and confirmed for scoring the importance of TPP1 in telosome assembly KO efficiency by western blotting (Figure 2a). The cells and POT1 telomere targeting [24, 25]. Notably, POT1 were then harvested for telomere chromatin immuno- KO also significantly reduced the targeting of both precipitation (ChIP) assays (Supplementary TPP1 and TIN2 to telomeres. Taken together, these Figure S3A). data suggest that TIN2, TPP1 and POT1 may form a Of the six proteins, both TRF1 and TRF2 can bind tight subcomplex. It is also clear that with the exception double-stranded telomeric DNA [9–13], and as expec- of RAP1, knocking out any of the subunits had a more ted, we found that TRF2 KO had no effect on TRF1 global effect on the remaining subunits (Figure 2b), binding to telomeres (Figure 2b). Using ectopically indicating significant contribution of each protein to expressed proteins, we showed previously that TIN2 the proper assembly of the telosome complex and that was essential for telosome assembly [23]. Although their roles in maintaining telosome function may be POT1 can bind ss telomeric DNA [14, 15], targeting of more complex than previously surmised. POT1 to telomeres requires TPP1, which in turn is tethered to telomeres through TIN2 [23, 24]. Consistent Activation of telomere DDRs in inducible KO cell lines with these previous findings, reductions in telomere Considerable efforts have been devoted to delineat- targeting of the remaining subunits were apparent in ing the complex signaling pathways that protect telo- TIN2 KO cells, with TPP1 and POT1 being the most meres and prevent the activation of DDR. Disruption affected (Figure 2b). Similarly, knocking out TPP1 also of the telosome complex can expose telomere ends to led to drastic reductions in telomere ChIP signals for the DDR machinery and eventually lead to

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 5

Figure 3 Deletion of individual subunits impacts telomere end protection. The inducible KO cell lines were maintained in the presence (+) or absence (−) of doxycycline for 6 days before being collected and used in the following assays. (a) The cells were examined by IF-fluorescence in situ hybridization (FISH) using antibodies against 53BP1 and the respective targeted proteins along with a telomere PNA probe. 4,6-Diamidino-2-phenylindole (DAPI) was used to stain the nuclei. Scale bars 10 μm. (b) Data from a were quantified and graphed. Three experiments with independent doxycylcine inductions were carried out for each cell line, with at least 100 cells analyzed in each experiment. Only cells with at least five TIFs were counted as positive. Error bars indicate s.e. (n = 3). P-values were obtained using the Student’s t-test. ***Po0.001. (c) The cells were harvested for metaphase spread and FISH analysis using a telomere probe. Representative images are shown here for cells induced to KO TIN2 and TRF2. White arrows indicate chromosome fusions. Scale bars 5 μm. (d) Data from c were quantified and graphed as indicated. At least 50 metaphases were scored for each sample. White arrowheads indicate chromosome fusions. Scale bars 5 μm. P-values were determined by one-way analysis of variance (ANOVA). (e) The cells were immunoblotted using the indicated antibodies. The anti-SMC1 antibody served as a loading control. γ-irradiated Hela cells (IR+) served as positive controls. (f) Data from e were quantified. At least three experiments with independent doxycycline inductions were performed and the results were combined. Signals for phosphorylated Chk1 and Chk2 were normalized against SMC1 signals and graphed as indicated. Error bars indicate s.e. P-values were obtained using the Student’s t-test. *Po0.05, **Po0.01. chromosomal abnormalities and cell cycle arrest [8]. marker. In addition, KO of TRF2 and TIN2 resulted in Immunofluorescence (IF) analysis of these inducible a marked increase in telomere fusions (Figure 3c and d, KO cell lines supports previous findings of the impor- Supplementary Figure S3B). These results again rein- tance of core telomeric proteins to telomere protection. force the notion that the six telomeric proteins have As evidenced by the recruitment of 53BP1 to telomere distinct roles in end protection and genomic stability. dysfunction induced foci (TIFs) (Figure 3a and b), Increased DDRs at telomeres can lead to activation upon doxycycline-induced KO, activation of DDRs at of ataxia-telangiectasia mutated (ATM) and ATM- telomeres could be observed. Except for RAP1 KO and Rad3-related (ATR) signaling, and the subsequent cells, which displayed minimal increase in TIFs, all phosphorylation of Chk2 and Chk1, respectively. other KO cell lines exhibited significant increases in TRF2 dysfunction has been shown to activate ATM 53BP1 foci that co-stained with a telomere DNA pathways [58, 59], whereas the POT1-TPP1

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 6

Figure 4 Deletion of individual subunits impacts telomere length and overhang maintenance. The inducible KO cell lines were maintained in the presence (+) or absence (−) of doxycycline (Dox) and collected at different time points for the following assays. At least three experiments with independent doxycycline inductions were performed and the results were combined. (a, b) Genomic DNA was extracted from the cells for telomere restriction fragment (TRF) analysis using a 32P-labeled telomere

probe (TTAGGG)3. Telomere signals were quantified and processed using TeloRun, and average telomere length was calculated and graphed for each cell line in a. Representative gels of the TRF assay are shown in b.(c) The cells were collected 6 days after induction and immunostained using antibodies against each targeted protein and RPA1 along with a telomere PNA probe. 4,6- Diamidino-2-phenylindole (DAPI) was used to stain the nuclei. Three independent experiments were carried out with at least 100 cells examined for each experiment. Scale bars 10 μm. (d) The cells were harvested 6 days after induction for genomic DNA extraction. The DNA was then processed in the presence (+) or absence (−) of Exonuclease I (ExoI) for in-gel hybridization 32 analysis of ss G overhangs. G overhangs were detected in the native gel using the P-labeled (CCCTAA)3 probe. Total telomeric DNA and Alu repeat signals were determined under denaturing conditions. (e) Overhang signals for each cell line from d were quantified and normalized against Alu repeat signals. Results from doxycycline-treated samples were compared with untreated samples and graphed as indicated. At least three independent experiments were performed for each cell line. Error bars indicate s.e. (n = 3). P-values were obtained using the Student’s t-test. **Po0.01.

heterodimer is important for inhibiting ATR activation somewhat unexpected, because deletion of mouse [60, 61]. Indeed, marked induction of phosphorylation Pot1a mainly induced Chk1 activation and Pot1b of Chk2 upon TRF2 KO and Chk1 upon POT1/TPP1 inactivation mostly impacted telomere overhangs KO was observed (Figure 3e and f, Supplementary [60–65]. Perhaps Chk2 activation in our POT1 KO cells Figure S4). In comparison, RAP1 deletion had no was a result of reduced telomere-associated TRF2 and impact, whereas TPP1 and TRF1 appear to be more TIN2 upon POT1 deletion. These results further specific for ATR-mediated DDR regulation. TIN2 and highlight the complex mechanisms that are in place to POT1 are both important for DDR, and their KO protect telomeres from DDR and the distinct signaling resulted in robust phosphorylation of both Chk2 and events mediated by each subunit, and suggest that more Chk1. The Chk2 response in POT1 KO cells was functional differences may exist between human and

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 7 mouse telomeric proteins in checkpoint response than recruitment of RPA1 could be observed upon deletion previously thought. of TRF2, TRF1 or RAP1 (Figure 4c and Supplementary Figure S6). It is possible that TRF1 and Telomere length maintenance and telomere overhang TRF2 can each independently bring the TPP1–POT1 protection in inducible KO cells complex to telomeres to protect G overhangs. Each of the six telomeric proteins participates in the In mice, ablation of Tin2, Tpp1 or Pot1a/b led to regulation of telomere length. Previous assessment of extended overhang length [48, 62, 63, 78–81]. Surpris- their roles in telomere length control has mostly relied ingly, of the six KO lines, only cells induced to KO on RNAi KD and overexpression of mutant proteins in POT1 exhibited an increase in overhangs (Figure 4d human cells, which has sometimes yielded conflicting and e). We found extensive chromosomal fusions upon results. For example, overexpression and RNAi TFR2 KO (Figure 3c and d), which likely compro- experiments indicate that TPP1 and POT1 negatively mised overhang protection and caused the slight regulate telomere length [21, 22, 24, 29, 66–68], but decrease in G overhang length in TFR2 KO cells. disrupting the TEL patch (TPP1 glutamate (E) and Overlapping phenotypes in mouse cells knocked out of leucine (L)-rich patch) within TPP1 led to decreased Tin2, Tpp1 or Pot1a/b, such as TIF induction and telomere length [69], the latter consistent with the overhang elongation, underline the interdependence of positive role TPP1 has in recruiting and promoting these proteins. The unexpected lack of overhang elon- telomerase activity [25, 26, 70, 71]. In this study, we gation in our TIN2 and TPP1 KO cells suggests that sought to better understand how inactivating indivi- POT1 may have a protection function independent of dual telomeric proteins may impact telomere length TIN2 and TPP1 in human cells, a major difference control using the inducible KO cells. between mouse and human cells in overhang Deleting the telomere duplex binding proteins TRF1 regulation. and TRF2 resulted in significantly elongated telomeres within a few days following doxycycline addition Human POT1 isoforms participate in telomere overhang (Figure 4a and b, Supplementary Figure S5). In the regulation case of TRF2, increased telomere fusions following The KO cell lines offer a unique opportunity for us induced KO (Figure 3c and d) likely led to the apparent to investigate the possible functional significance of rapid increase in telomere length observed here. In splicing variants of telosome subunits. Although comparison, inducible RAP1, TIN2 and POT1 KO human has one POT1 gene as opposed to two in mice, a cells showed a more gradual increase in telomere total of five alternatively spliced forms of hPOT1 have length. The TPP1 KO cells exhibited moderate accel- been described to date [15]. hPOT1 V1 is the full-length eration of telomere shortening, which is more in line form that has been extensively studied (Supplementary with TPP1’s role in recruiting and promoting telo- Figure S7A). Little is known about the functional sig- merase activity. nificance of the other isoforms, which appear to be Mammalian telomeres are thought to adopt the expressed in normal and cancer tissues as well as cancer t-loop structure, with the ss G overhang invading into cell lines [15, 82]. The V4 variant was not examined the duplex DNA [72–74]. The G overhangs are main- here because POT1 coding sequences are interrupted tained through telomere DNA synthesis and active by an early stop codon. The remaining isoforms share resection by exonucleases of the C-strand on 5′ ends with V1 the N-terminal OB-fold domain but differ in [4, 8]. Evidence suggests that the POT1–TPP1 complex their C-termini. We have designed the dual sgRNAs to coats telomere G overhangs, inhibits nucleolytic inactivate all of the POT1 isoforms (Supplementary attacks and prevents the binding of the nonspecific Figure S7A), enabling us to determine the role of each ssDNA binding protein RPA1 and subsequent activa- isoform individually. We expressed CRISPR-resistant tion of ATR-mediated checkpoint responses [29, 30, POT1 V1 isoform in the POT1 KO cells and examined 75–78]. Furthermore, TIN2 deletion in mice also led to the cells for overhang status, TIF formation and RPA1 RPA1 accumulation and ATR activation, in line with accumulation (Figure 5 and Supplementary Figure its role in tethering POT1–TPP1 to telomeres [78]. S7B–D). As expected, POT1 V1 could localize to tel- Consistent with these previous findings, inhibition of omeres (Figure 5b), and rescue the phenotypes of TPP1, POT1 and TIN2 led to aberrant accumulation increased TIFs, accumulated RPA1, and excessively of RPA1 at telomeres in 13.2%, 21.7% and 12.5% of the long overhangs (Figure 5c–f). These data support V1 as cells, respectively (Figure 4c), indicating deprotected the main POT1 isoform that caps telomere ends and G overhangs. In contrast, no upregulated telomeric shields them from DDRs.

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 8

Figure 5 Human POT1 variants have important roles in maintaining telomere overhangs. (a) 293T cells transiently co-expressing GST-tagged TPP1 and Flag-tagged human POT1 variants were immunoprecipitated (IP) with anti-Flag antibodies and blotted as indicated. In addition to wild-type (wt) POT1 variant proteins, rescue constructs encoding POT1 variants that contained sgRNA- resistant silent mutations (rescue) were also examined. POT1 KO cells stably expressing the sgRNA-resistant rescue constructs were then induced with doxycycline for 6 days before being harvested for the assays described below. (b) For telomere ChIP analysis, the rescue cells were immunoprecipitated (IP) with anti-HA antibodies to bring down associated telomeric DNA for dot-

blotting and hybridization with a telomere repeat probe (TTAGGG)3 (left panel). IgG was used as a negative control. Telomeric signals were quantified and normalized against Alu repeat signals and graphed on the right. Error bars indicate s.e. P-values were obtained using one-way analysis of variance (ANOVA). *Po0.05, **Po0.01. (c) To assess RPA1 recruitment to telomeres, cells were immunostained with antibodies against RPA1 along with a telomere PNA probe. The data were quantified and graphed as shown. Three independent experiments were carried out for each cell line with at least 100 cells in each experiment. Error bars indicate s.e. (n = 3). P-values were obtained using the Student’s t-test. **Po0.01, ***Po0.001. (d) To assess possible changes in TIFs, cells were immunostained using antibodies against 53BP1 along with a telomere PNA probe. The data were quantified and graphed as shown. Three independent experiments were carried out for each cell line with at least 100 cells in each experiment. Only cells with at least five TIFs were counted as positive. Error bars indicate s.e. (n = 3). P-values were obtained using the Student’s t-test. *Po0.05, ***Po0.001. (e) Genomic DNA was extracted from the cells for processing in the presence

(+) or absence (−) of Exonuclease I (ExoI) before hybridization analysis of ss G overhangs in the native gel using the (CCCTAA)3 probe. Total telomeric DNA and genomic DNA (Alu) signals were determined under denaturing conditions. (f) Overhang signals from e were quantified and normalized against Alu repeat signals and graphed as indicated. At least three independent experiments were performed for each cell line. Error bars indicate s.e. (n = 3). P-values were calculated using one-way analysis of variance (ANOVA).

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 9

Figure 6 Deletion of individual subunits impacts metabolic pathways in the inducible KO cells. (a) Core metabolic pathways and key metabolites in mammalian cells are highlighted. G6P, glucose-6-phosphate. F6P, fructose-6-phosphate. FBP, fructose-1,6,- biphosphate. 3PG/2PG, 3-phosphoglycerate/2-phsphoglycerate. PEP, phosphoenolpyruvate. OAA, oxaloacetate. (b, c) Inducible KO cells were cultured in the presence of doxycycline for 6 days before being collected in replicates and processed for targeted metabolomic analysis by LC-MS. Cas9-inducible parental cells were grown in the presence of doxycycline and used as controls (Con). The results were normalized to internal standards and metabolites that consistently showed differences in two experiments with independent doxycycline inductions are presented here. Error bars correspond to s.d. of three independent experiments. P-values were calculated using one-way analysis of variance (ANOVA). *Po0.05; **Po0.01; ***Po0.001. (d) The six telomere proteins may assemble and function on telomeres as a single unit (1), a TRF1-less five-protein subcomplex (2), or a four-protein subcomplex without TRF2/RAP1 (3). Non-telomere bound subcomplexes also exist (4). (e)A model of telomere protection that incorporates human POT1 isoforms highlights the unique features of the human system.

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 10 Consistent with their lack of TPP1-interacting bioenergetic pathways to support growth of cancer domains, POT1 variants V2, V3 and V5 could not cells in which upregulated glycolysis and glutamino- interact with TPP1 (Figure 5a). All three isoforms do lysis pathways have often been found. contain intact OB-folds that can mediate telomere Given that treatment of cells with drugs such as ssDNA binding, and therefore may target to telomeres doxycycline can drastically alter cellular metabolomes, independent of TPP1. In support of this notion, our we decided to compare doxycycline-induced TIN2 KO telomere ChIP experiments showed that POT1 V2, V3 cells with doxycycline-treated Cas9-inducible parental and V5 could indeed associate with telomeres cells that did not express any sgRNA sequences. The (Figure 5b and Supplementary Figure S7B), albeit with inducible TIN2 KO cells were expanded and then markedly reduced abilities compared with V1. Binding treated with doxycycline for 6 days before being col- of POT1 isoforms to telomeres appeared to increase in lected for analysis by quantitative liquid chromato- the absence of endogenous POT1 (Figure 5b), probably graphy–mass spectrometry (LC-MS). As shown in because of more sites becoming available upon POT1 Figure 6b, TIN2 KO led to varying changes in a broad deletion, or because of increased overhang length. range of metabolites in glycolysis, TCA cycle and Individual expression of the short POT1 isoforms in macromolecule synthesis. When we examined the other POT1 KO cells could not rescue the phenotypes of telosome subunits, we found TRF1 KO to have the RPA1 accumulation or increased telomere DNA least impact, only consistent increases in ribose (data damage (Figure 5c and d and Supplementary Figure not shown). In comparison, TRF2 KO led to repro- S7C and D). However, they consistently reduced the ducible increases in a number of metabolites increase in telomere overhangs in these cells, albeit to (Figure 6c). Similarly, knocking down the remaining varying degrees (Figure 5e and f). These data indicate subunits resulted in reproducible and differential that the shorter human POT1 isoforms can regulate changes in certain key metabolites in glycolysis and overhang length independent of TPP1. Furthermore, macromolecule synthesis. These findings reaffirm the this previously unknown function of the POT1 iso- distinct roles that each subunit has in ensuring the forms may help to explain some of the differential growth and proliferation of the cell. None of the other phenotypes observed between mouse and human proteins examined had the same widespread effect on KO cells. metabolism as TIN2. For example, TIN2 was the only telosome subunit whose KO affected multiple meta- Deletion of telosome subunits leads to metabolic bolites in the TCA cycle, which occurs in the mito- perturbations in the inducible KO cells chondria. This observation supports our previous Complete deletion of essential genes such as TRF2 findings that TIN2 can localize to the mitochondria and TIN2 causes cell cycle arrest and/or death, making and regulate the metabolic pathways in the it difficult to isolate single KO clones or obtain large mitochondria. numbers of cells for extensive biochemical studies. Our inducible KO system bypasses the need of KO cell Discussion cloning, and enables the expansion of cells to large quantities before KO induction for biochemical ana- Work using mouse models, mutant proteins and lysis such as metabolomic profiling. RNAi to probe the functional significance of the telo- Crosstalk between telomere maintenance and some and its subunits has greatly advanced our metabolic pathways has been well documented, with knowledge and understanding of telomere home- several key telomere regulators including TERT and ostasis. Genetically modified mouse models have been TIN2 implicated in more direct metabolic regulation indispensable to loss-of-function studies, but differ- [33–37, 83]. We therefore examined TIN2 KO cells and ences between mouse and human, as well as the cost determined how completely disrupting TIN2 affected and efforts associated with mouse studies continue to metabolic control, especially with respect to key pose challenges. Major drawbacks of RNAi-mediated metabolites in glycolysis and the tricarboxylic acid KD include its off-target effects and the inability to (TCA) cycle. Research has shown that cancer cells achieve complete inhibition. As a result, many ques- often consume large quantities of glucose, which fuels tions regarding human telomere maintenance remain the TCA cycle, as well as pathways for macromolecule unanswered. The RNA-guided CRISPR/Cas9 synthesis (for example, nucleotides, amino acids and genome-editing technology has enabled unprecedented lipids) (Figure 6a) [84, 85]. Glucose and other meta- manipulations of the genome in a much more targeted bolites such as glutamine serve as substrates in various and efficient manner, particularly in somatic cells and

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 11 cell lines [39–47]. In this report, we describe the sys- they can still associate with telomeres (Figure 5a and tematic generation and profiling of inducible KO cell b). Given that our POT1 KO strategy disrupts isoforms lines for the six core telomere proteins. In all of the V2, V3 and V5 as well, we speculate that these POT1 experiments presented, the results came from multiple isoforms may participate in overhang protection inde- independent doxycycline inductions of the inducible pendent of TPP1 and that the unexpected results seen KO cells. Such reproducibility and consistence under- in TIN2 and TPP1 KO cells help to highlight this score the robustness of the inducible system and the shared function between different human POT1 iso- advantage of using polyclonal populations in cellular forms (Figure 6e). assays. We can now more clearly define the function of Based on this model, telomere targeting of the full- human telomeric proteins and identify differential length variant POT1 V1 is disrupted in TPP1 and TIN2 regulatory mechanisms in human vs mouse. KO cells; however, telomere ssDNA overhangs can still be protected by other variants, which can localize Organization of the human telosome/shelterin complex to telomeres independent of TPP1-TIN2. These OB- With the inducible KO system, we now have a fold only POT1 proteins may associate with telomeres clearer picture of how the human telosome complex directly or through interaction with other OB-fold may be organized (Figure 6c). Of the six subunits, containing proteins. Indeed, when we ectopically RAP1 only interacts with TRF2. Except for telomere expressed V2, V3 or V5 in the POT1 KO cells, they length changes, induced RAP1 deletion had no major could rescue the overhang length phenotype to varying impact in most of the telomere assays described here, degrees. Similar findings were previously reported for consistent with previous analyses using RAP1- POT1 V5 in POT1 KD cells [82]. Interestingly, POT1 inactivated mouse and cellular models [86–88]. In V2, V3 and V5 could not rescue the TIF or RPA1 addition, removing either TIN2, TPP1 or POT1 phenotypes of POT1 KO cells, suggesting distinct markedly impacted the telomere targeting of the other pathways for different POT1 isoforms in regulating two proteins, providing strong evidence that these three overhang length vs DDR. It is possible that once V2, proteins likely form a functional unit on telomeres. V3 and V5 are recruited to telomeres, they can block Interestingly, TRF2 KO affected the telomeric binding exonucleases such as Exo1 from further recessing the 3’ of all the other subunits except for TRF1, supporting end of telomeres. Collectively, our study supports a the existence of the five-protein TRF1-less complex new model of both TPP1-dependent and -independent (Figure 6d). regulation of telomere overhangs by human POT1, in contrast to the TPP1-dependent model for mouse The role of POT1 in G overhang protection Pot1a/b. In mice, both Pot1a and Pot1b can bind Tpp1 and are tethered to telomeres through Tpp1-Tin2; however, The role of telomeric proteins in metabolic control the two mouse POT1 proteins participate in distinct Although previous studies have implicated telomeric signaling events for telomere regulation [68, 89]. proteins in metabolic regulation, this is the first time Human POT1, in comparison, appears to carry out the that metabolic changes were systematically investi- functions attributed to both POT1a and POT1b. It is gated upon deletion of individual subunits of the shel- therefore expected that depletion of TPP1 or TIN2 in terin/telosome complex. It is possible that the human cells would disrupt POT1 targeting to telomeres metabolic alterations observed in our KO cells were and POT1-mediated protection and length regulation indirect results of activation of DDR pathways and of G overhangs. For example, Tpp1 KO MEFs showed changes in telomere length. However, the differences in similar phenotypes as cells doubly knocked out for the metabolomes in these cells suggest distinct and Pot1a/b [79, 80]. Although POT1 KO in our inducible significant impact on cellular metabolism as a result of cells led to expected increases in overhang length and inhibition of different human telomeric proteins. For RPA1 staining, we were surprised to discover a lack of instance, TIN2 KO appeared to impact many of the accumulation of elongated telomere ssDNA in cells metabolites in different metabolic pathways, including knocked out of TIN2 or TPP1. Of the four splicing the TCA cycle. The TCA cycle, which occurs in the variants of human POT1 examined in this study, only mitochondria where it metabolizes the end products of V1 can interact with TPP1, likely because it is the only glycolysis and feeds into oxidative phosphorylation, is variant that retains the C-terminal TPP1-interacting central to energy production and biosynthesis. The domain (Figure 5a and Supplementary Figure S7A). finding that only TIN2 KO appeared to impact TCA Although isoforms V2, V3 and V5 do not bind TPP1, cycle metabolites supports our previous findings of

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 12 TIN2 targeting to the mitochondria, and is consistent induced cells are not single clones, possible complica- with idea that TIN2 can directly regulate metabolism. tions in data interpretation from off-target effects may Interestingly, although TRF2 KO appeared to also be less likely. Moreover, expression of CRISPR- affect multiple glycolytic, glutaminolytic and nucleic resistant constructs could rescue the observed pheno- acid synthesis intermediates, knocking out the types, which helped to rule out potential off-target remaining subunits was more restricted in terms of effects. changes in the metabolome. Whether such differences It appeared to take similar amount of time for our are linked to the predominant function of TRF2 in inducible cells to achieve efficient KO as for cells ATM-mediated DDR response warrants further transfected with plasmids encoding Cas9 and sgRNA investigation. Taken together, these data underline the to generate straight KO. The isolation of straight KO complex crosstalk between telomere maintenance and cells, however, requires significantly longer time (for metabolic control. non-essential genes), whereas large numbers of induced KO cells can be more quickly obtained. The inherently Application of the inducible CRISPR/Cas9 KO system mixed nature of the induced KO population does have For genes essential for growth and survival, the the potential to introduce variations. The distribution inducible KO cell lines afford the time window needed of various populations (+/+ vs +/ − vs − / − ) should in to carry out biochemical studies before the cells theory be similar each time the cells are induced. Deep undergo growth arrest. For example, we could not sequencing of the induced KO cells may help to shed obtain straight KO clones of cells deleted for TIN2, light on the exact dynamics of the KO populations, but TRF2 or POT1, but the inducible KO cells have the short reads of such methods will fail to capture any allowed us to explore the functions of these proteins in large deletions, especially with our dual sgRNA cells. a variety of assays. In theory, a single sgRNA targeting Any variability in population dynamics is more likely a specific site within a locus should effectively generate to be caused by variable Cas9 induction than the con- cells with frame-shift indels that inactivate the target stitutively expressed sgRNAs. Hence, using the same gene. In our induced single sgRNA-targeted KO cells, inducible Cas9 clone to generate all six cell lines for we often found low expression levels of the target genes comparative studies should help minimize variations. after Cas9 induction. This residual expression is likely a Furthermore, each of our cell lines had been induced result of the induced KO cells containing a mixture of multiple times independently for each assay and the alleles with different indels (some of which cannot results were found to be similar, attesting to the completely disrupt target gene function), and differs reproducibility of the system. from that observed in RNAi KD cells. In the latter, The six inducible cell lines should prove particularly every cell likely still expresses the target gene at a cer- useful to investigators who may be interested in tain level following incomplete suppression by small studying different aspects of telomere maintenance. interfering RNAs/short hairpin RNAs. The polyclonal For instance, in addition to the ability to assay cells in induced KO cells, on the other hand, comprise mostly which essential genes are deleted, this system also of cells completely knocked out for the target gene, enables real-time comparisons of specific protein with a small fraction of cells that may contain in-frame complexes before and after the removal of a key sub- indels or can ‘restore’ expression following extended unit. Furthermore, our method makes possible both culturing. This is an important distinction because the production of large numbers of genetically edited residual expression in nearly every RNAi KD cell may cells for studies of essential genes and the generation of be sufficient for the entire population to ‘behave’ nor- more well-defined snapshots of cells in response to mally in an assay. However, a very small fraction of telomere perturbations. cells with heterozygous KO or wild-type alleles are unlikely to dilute or mask the response of the whole Materials and Methods population, if the overwhelming majority of cells have no expression of the gene. Vector construction Based on our data, more complete deletion and Sequences encoding the humanized Cas9 gene under the inactivation can be achieved with two sgRNAs. A control of the tetracycline-responsive promoter were cloned into fi a lentiviral vector that also encodes rtTA [90]. Individual second sgRNA signi cantly reduced the possibility of sgRNA sequences (Supplementary Table S1) were cloned into in-frame ligations, as well as the ability of cells to modified vectors encoding different antibiotic resistance genes overcome inactivating mutations, although with the (puromycin, blasticidin or hygromycin). These vectors are based caveat of possibly increasing off-targets. As the on the LentiCRISPR vector (GeCKO) but no longer contain

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 13

Cas9 sequences [90]. Complementary encoding wild-type anti-TPP1 and anti-TIN2 polyclonal antibodies [6], and goat and rescue mutants for sgRNA-resistant hTIN2 and hPOT1 anti-TRF1 antibody [23], rabbit anti-p-Chk1(Ser317) and anti- isoforms were cloned into a pHAGE-based lentiviral vector for Chk1 antibodies (Cell Signaling Technology, Danvers, MA, C-terminal tagging with HA and FLAG epitopes [91]. Rescue USA), and rabbit anti-p-Chk2(Thr68) and anti-Chk2 antibodies constructs were generated by introducing either single- (Cell Signaling), mouse anti-p-ATM (Ser1981) and rabbit anti- nucleotide mutations in the PAM sequence (TIN2 sgRNA1: ATM antibodies (Cell Signaling), rabbit anti-p-ATR(Ser428) GGG → GAG, TIN2 sgRNA2: TGG → TGA, POT1 sgRNA1: and anti-ATR antibodies (Cell Signaling) and rabbit anti-HA AGG → AGA) or double-nucleotide mutations in the sgRNA antibody (Santa Cruz Biotechnology). target region (POT1 sgRNA2: GGAGGTACCAGTTAC → GGTCG GGAGGTACCAGTTACGGAAG). Cell proliferation assay and cell cycle analysis Cells were plated in 12-well plates at 1 × 104 cells per well and Generation of inducible CRISPR KO cell lines maintained for 10 days with or without 1 μgml−1 doxycycline. Hela cells stably expressing doxycycline-inducible Cas9 were The number of viable cells at various time points was determined first generated by lentiviral transduction. A single clone with by Trypan blue exclusion. To determine DNA content, 1 × 106 robust and efficient Cas9 induction was selected for further cells maintained with or without doxycycline (1 μgml−1) for experiments. Vectors expressing single or dual sgRNAs were 6 days were collected, washed with 1 × phosphate-buffered sal- stably introduced into the Cas9-inducible cells by lentiviral ine and then fixed in 70% ethanol at room temperature for transduction followed by selection with appropriate antibiotics. 30 min. The fixed cells were then incubated in 0.5 ml 1 × phos- The appropriate concentrations for Cas9 induction and efficient phate-buffered saline containing 50 μgml−1 propidium iodide cleavage at the intended locus were determined for each cell line. and 0.2 mg ml−1 DNase-free RNase A (pH 7.4) at 37 °C for We have found that 6 days of incubation in 1 μgml−1 of dox- 30 min. The cells were subsequently analyzed using an LSRII ycycline is optimal for our cells. Successful inactivation of each flow cytometry analyzer (BD Biosciences, San Jose, CA, USA). gene was confirmed by western blotting with the appropriate antibodies. Further validation was conducted by extracting IF and telomere fluorescence in situ hybridization (FISH) genomic DNA from the cells either for direct sequencing or for analysis TOPO cloning before Sanger sequencing. IF was performed as previously described [25]. Briefly, cells grown on glass coverslips were permeabilized for 30 s with 0.2% Immunoprecipitation, immunoblotting and telomere ChIP Triton X-100, fixed in 4% paraformaldehyde and then permea- assays bilized again with 0.5% Triton X-100, before being blocked in Co-immunoprecipitation studies were performed as descri- 5% bovine serum albumin. Cells were subsequently incubated bed previously [19]. Cells were lysed in 1 × NETN buffer (1 M with appropriate antibodies and/or a telomere peptide nucleic Tris-HCl (pH 8.0), 1 mM EDTA, 100 mM NaCl and 0.5% acid (PNA) probe (Bio-PNA). 4,6-Diamidino-2-phenylindole Nonidet P-40) containing 1 mM DTT and a proteinase inhibitor was used to visualize the nuclei. For TIF assays, 4100 cells were mixture (Roche Applied Science, Mannheim, Germany). The examined for each experiment, and cells with 45 co-stained foci lysates were then immunoprecipitated with appropriate anti- were counted as being TIF positive. At least three independent bodies for sodium dodecyl sulfate–polyacrylamide gel electro- experiments were performed for each cell line. phoresis and western blotting. Antibodies used for IF are: mouse anti-FLAG M2 and rabbit Telomere ChIP assays were performed as described pre- anti-FLAG polyclonal antibodies (Sigma), rabbit anti-53BP1 viously with slight modifications [92]. Briefly, cells were chemi- (NB100-304; Novus Biologicals) and mouse anti-53BP1 (BD cally crosslinked in1% formaldehyde in phosphate-buffered Biosciences) antibodies, rat anti-RPA1 polyclonal antibody saline, and sonicated to shear chromatin. Sonicated lysates were (Cell Signaling), goat anti-TRF1 polyclonal antibody, mouse pre-cleared before being incubated with 3 μg of antibodies for anti-TRF2 monoclonal antibody (Calbiochem), rabbit anti- immunoprecipitation. The co-precipitated DNA was eluted and RAP1 polyclonal antibody (Bethyl Laboratories), rabbit anti- analyzed by dot-blot and southern hybridization using the 32P- TPP1 and anti-TIN2 polyclonal antibodies [6], rabbit anti-POT1 labeled telomere (TTAGGG)3 and Alu repeat probes. polyclonal antibody (Bethyl Laboratories) and rabbit anti-HA Antibodies used for immunoprecipitation and western blot antibody (Santa Cruz Biotechnology). analyses in this study are: horseradish peroxidase-conjugated Metaphase spread and telomere fluorescence in situ hybridi- anti-glutathione S-transferase (GST) polyclonal antibody (GE zation analysis was performed as previously described [93]. Healthcare Life Science, Pittsburgh, PA, USA), horseradish Briefly, cells were incubated with 0.1 μgml−1 colcemid (Kar- peroxidase-conjugated anti-FLAG M2 antibody and M2- yoMax, Invitrogen, Carlsbad, CA, USA) for 3 h and harvested. conjugated agarose beads (Sigma, St Louis, MO, USA), rabbit The cells were then incubated in hypotonic solution (0.075 M anti-FLAG polyclonal antibody (Sigma), goat anti-actin poly- KCl) for 25 min at room temperature, fixed in methanol/glacial clonal antibody (Santa Cruz Biotechnology, Dallas, TX, USA) acetic acid (3:1) solution for 5 min and spread onto clean slides. and rabbit anti-SMC1 antibody (Bethyl Laboratories, Mon- The slides were treated with pepsin, fixed in 4% formaldehyde, tgometry, TX, USA), mouse anti-TRF2 monoclonal antibody dehydrated in successive ethanol baths (70, 90, and 100%) for (Calbiochem, San Diego, CA, USA), rabbit anti-RAP1 poly- 5 min each and air dried. The slides were subsequently dena- clonal antibody (Bethyl Laboratories), rabbit anti-POT1 poly- tured at 80 °C for 3 min and then hybridized with telomere PNA clonal antibody (Novus Biologicals, Littleton, CO, USA), rabbit probes (Bio-PNA) at 25 °C for 2 h in the dark. The slides were

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 14

then washed, dehydrated and mounted with VectaShield normal phase chromatographic separation. All samples were mounting medium containing 4,6-diamidino-2-phenylindole kept at 4 °C and 5 μl was used for analysis. (Vector Laboratories, Burlingame, CA, USA). At least 50 TCA metabolites were separated through normal phase metaphase spreads were captured using a Zeiss Imager Z1 chromatography. The binary pump flow rate was 0.2 ml min−1 microscope (Göttingen, Germany) and analyzed using AxioVi- with 80% B to 2% B gradient over 20 min, 2% B to 80% B for sion 4.8(Göttingen, Germany). 5 min and 80% B for 13 min. The flow rate was gradually increased as follows: 0.2 ml min−1 (0–20 min), 0.3 ml min−1 (20.1–25 min), 0.35 ml min−1 (25–30 min), 0.4 ml min−1 TRF assay and telomere overhang analysis (30–37.99 min) and 0.2 ml min−1 (5 min). Metabolites were Cells were first induced with doxycycline for 6 days, and then separated on a Luna Amino (NH2) column (4 μm, 100A collected for TRF analysis at various time points to estimate the 2.1 × 150 mm, Phenominex, Torrance, CA, USA) in a average length of telomeres using TeloRun [6, 94]. The in-gel temperature-controlled chamber (37 °C). All columns used were detection of telomere ssDNA overhangs was performed as washed and reconditioned after every 50 injections. HPLC- previously described with slight modifications [95]. Genomic grade acetonitrile, methanol and water were from Burdick & DNA was digested with HinfI and RsaI (New England Biolabs, Jackson (Morristown, NJ, USA). The calibration solution Ipswich, MA, USA) for 16 h. In all, 5 μg each of the digested containing multiple calibrants in acetonitrile/trifluroacetic acid/ DNA was then incubated with and without 20 U of Exonuclease water was from Agilent Technologies. Data were curated I (New England Biolabs) for 6 h. The reaction mixtures were through quality control assessment (using data from sample fractionated on a 1.2% agarose gel in 1 × Tris/borate/EDATA pools) and normalization using internal standards. (TBE) buffer for 1.5 h at 60 V and dried on a gel dryer at 50 °C for 2 h. The dried gel was pre-hybridized in hybridization buffer (0.5 M Na2HPO4 pH 7.2, 1 mM EDTA, 7% sodium dodecyl Statistical analysis sulfate) and hybridized in fresh buffer with a 32P-labeled All experiments were independently repeated at least three times and presented as mean ± s.d. or mean ± s.e. Statistical C-strand probe (CCCTAA)3. The gel was washed 3 × with 2 × SSC containing 0.1% sodium dodecyl sulfate for 30 min at analyses were performed using either Student’s t-test or one-way room temperature and analyzed on a PhosphorImager (GE analysis of variance. Significant differences were defined as Healthcare Life Science). The gel was subsequently denatured Po0.05 or lower. (0.5 M NaOH, 1.5 M NaCl for 30 min), neutralized (0.5 MTris- HCl for 30 min), and then hybridized with the C-strand or Alu fl repeat probes as control. Con ict of Interest

The authors declare no conflict of interest. Metabolomic analysis by liquid chromatography-mass spectrometry Each cell line was induced with 1 μgml−1 doxycycline, cul- Acknowledgements tured and harvested in multiple (3–4) replicates (~5 × 106 cells each), and frozen in aliquots before metabolome extraction as This work was supported by the National Key Research and described previously [96]. Briefly, cells were three times frozen Development Program of China (2017YFA0102800 and and thawed and resuspended in ice-cold methanol:water (750 μl, 2017YFA0102801), National Natural Science Foundation of 4:1) containing 20 μl of internal standards (Tryptophan-15N2, China (NSFC 91640119 and 81330055), Science and Technol- Glutamic acid-d5, Thymine-d4, Gibberellic acid, Trans-Zeatin, ogy Planning Project of Guangdong Province Jasmonic acid, Anthranilic acid and Testosterone-d3, all from (2015B020228002) and Guangzhou Science and Technology Sigma-Aldrich). Homogenization entailed 2 × 30-s pulses, 10- Project (201605030012). We would also like to acknowledge the min vortex mixing with ice-cold chloroform (450 μl), and 2-min support of NIGMS GM095599, NCI CA211653, CPRIT vortex mixing with ice-cold water (150 μl). The homogenate was RP160462, the Welch Foundation Q-1673, HL131744 and the incubated at − 20 °C for 20 min and centrifuged at 4 °C for C-BASS Shared Resource (with special thanks to Dr JX) at the 10 min to partition aqueous and organic layers for drying at Dan L Duncan Cancer Center (DLDCC) of Baylor College of 37 °C for 45 min. The aqueous extract was reconstituted in Medicine (P30CA125123). This research was also made possible 500 μl of ice-cold methanol:water (50:50) and filtered at 4 °C for through the support by the CPRIT Core Facility Support 90 min through 3 kDa molecular filters (AmiconUltracel − 3K Award RP120092, Proteomic and Metabolomic Core Facility, Membrane, Millipore Corporation, Billerica, MA, USA). The NCI/2P30CA125123-09 Shared Resources Metabolomics core, filtrate was dried for 45 min at 37 °C before resuspension in and funds from the Dan L Duncan Cancer Center (DLDCC). 100 μl of methanol:water (50:50) containing 0.1% formic acid (Sigma-Aldrich). High-performance liquid chromatography (HPLC) analysis was performed using an Agilent 1290 series Author contributions HPLC system equipped with a degasser, binary pump, ther- mostatted autosampler and column oven (Agilent Technologies, HK, FL, QH, TD and FJ performed experiments; HK, FL, Santa Clara, CA, USA). The multiple reaction monitoring- QH and NP performed data analysis; HK, DL and ZS designed based measurement of relative metabolite levels used reverse or experiments; HK, DL and ZS wrote the manuscript.

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 15

References 21 Houghtaling BR, Cuttonaro L, Chang W, Smith S. A dynamic molecular link between the telomere length 1 Blackburn EH, Epel ES, Lin J. Human telomere biology: a regulator TRF1 and the chromosome end protector TRF2. contributory and interactive factor in aging, disease risks, Curr Biol 2004; 14: 1621–1631. and protection. Science 2015; 350: 1193–1198. 22 Ye JZ, Hockemeyer D, Krutchinsky AN et al. POT1- 2 Schmidt JC, Cech TR. Human telomerase: biogenesis, interacting protein PIP1: a telomere length regulator that trafficking, recruitment, and activation. Genes Dev 2015; recruits POT1 to the TIN2/TRF1 complex. Genes Dev 29: 1095–1105. 2004; 18: 1649–1654. 3 Arnoult N, Karlseder J. Complex interactions between the 23 O’Connor MS, Safari A, Xin H, Liu D, Songyang Z. DNA-damage response and mammalian telomeres. Nat A critical role for TPP1 and TIN2 interaction in high-order Struct Mol Biol 2015; 22: 859–866. telomeric complex assembly. Proc Natl Acad Sci USA 2006; 4 Armanios M, Blackburn EH. The telomere syndromes. Nat 103: 11874–11879. Rev Genet 2012; 13: 693–704. 24 Liu D, Safari A, O’Connor MS et al. PTOP interacts with 5 Hockemeyer D, Collins K. Control of telomerase action at POT1 and regulates its localization to telomeres. Nat Cell human telomeres. Nat Struct Mol Biol 2015; 22: 848–852. Biol 2004; 6: 673–680. 6 Liu D, O’Connor MS, Qin J, Songyang Z. Telosome, a 25 Xin H, Liu D, Wan M et al. TPP1 is a homologue of ciliate mammalian telomere-associated complex formed by mul- TEBP-beta and interacts with POT1 to recruit telomerase. tiple telomeric proteins. J Biol Chem 2004; 279: Nature 2007; 445: 559–562. 51338–51342. 26 Wang F, Podell ER, Zaug AJ et al. The POT1-TPP1 telo- 7 Xin H, Liu D, Songyang Z. The telosome/shelterin complex mere complex is a telomerase processivity factor. Nature and its functions. Genome Biol 2008; 9: 232. 2007; 445: 506–510. 8 Palm W, de Lange T. How shelterin protects mammalian 27 Zhong FL, Batista LF, Freund A et al. TPP1 OB-fold telomeres. Annu Rev Genet 2008; 42: 301–334. domain controls telomere maintenance by recruiting telo- 9 Bianchi A, Smith S, Chong L, Elias P, de Lange T. TRF1 is merase to chromosome ends. Cell 2012; 150: 481–494. a dimer and bends telomeric DNA. EMBO J 1997; 16: 28 Nandakumar J, Bell CF, Weidenfeld I et al. The TEL patch 1785–1794. of telomere protein TPP1 mediates telomerase recruitment 10 Broccoli D, Smogorzewska A, Chong L, de Lange T. and processivity. Nature 2012; 492: 285–289. Human telomeres contain two distinct Myb-related pro- 29 Loayza D, De Lange T. POT1 as a terminal transducer of teins, TRF1 and TRF2. Nat Genet 1997; 17: 231–235. TRF1 telomere length control. Nature 2003; 424: 11 Smith S, de Lange T. TRF1, a mammalian telomeric pro- 1013–1018. tein. Trends Genet 1997; 13:21–26. 30 Colgin LM, Baran K, Baumann P, Cech TR, Reddel RR. 12 van Steensel B, de Lange T. Control of telomere length by Human POT1 facilitates telomere elongation by telomer- the human telomeric protein TRF1. Nature 1997; 385: ase. Curr Biol 2003; 13: 942–946. 740–743. 31 Songyang Z, Liu D. Inside the mammalian telomere 13 Smogorzewska A, van Steensel B, Bianchi A et al. Control interactome: regulation and regulatory activities of of human telomere length by TRF1 and TRF2. Mol Cell telomeres. Crit Rev Eukaryot Gene Expr 2006; 16: Biol 2000; 20: 1659–1668. 103–118. 14 Baumann P, Cech TR. Pot1, the putative telomere end- 32 Robles-Espinoza CD, Velasco-Herrera Mdel C, binding protein in fission yeast and humans. Science 2001; Hayward NK, Adams DJ. Telomere-regulating genes and 292: 1171–1175. the telomere interactome in familial cancers. Mol Cancer 15 Baumann P, Podell E, Cech TR. Human Pot1 (protection Res 2015; 13: 211–222. of telomeres) protein: cytolocalization, gene structure, and 33 Santos JH, Meyer JN, Van Houten B. Mitochondrial alternative splicing. Mol Cell Biol 2002; 22: 8079–8087. localization of telomerase as a determinant for 16 Li B, Oestreich S, de Lange T. Identification of human hydrogen peroxide-induced mitochondrial DNA Rap1: implications for telomere evolution. Cell 2000; 101: damage and . Hum Mol Genet 2006; 15: 471–483. 1757–1768. 17 Kim SH, Kaminker P, Campisi J. TIN2, a new regulator of 34 Santos JH, Meyer JN, Skorvaga M, Annab LA, Van Houten B. telomere length in human cells. Nat Genet 1999; 23:405–412. Mitochondrial hTERT exacerbates free-radical-mediated 18 Ye JZ, Donigian JR, van Overbeek M et al. TIN2 binds mtDNA damage. Aging Cell 2004; 3:399–411. TRF1 and TRF2 simultaneously and stabilizes the TRF2 35 Santos JH, Hunakova L, Chen Y, Bortner C, complex on telomeres. J Biol Chem 2004; 16: 16. Van Houten B. Cell sorting experiments link persistent 19 Kim H, Lee OH, Xin H et al. TRF2 functions as a protein mitochondrial DNA damage with loss of mitochondrial hub and regulates telomere maintenance by recognizing membrane potential and apoptotic cell death. J Biol Chem specific peptide motifs. Nat Struct Mol Biol 2009; 16: 2003; 278: 1728–1734. 372–379. 36 Singhapol C, Pal D, Czapiewski R, Porika M, Nelson G, 20 Chen Y, Yang Y, van Overbeek M et al. A shared docking Saretzki GC. Mitochondrial telomerase protects cancer motif in TRF1 and TRF2 used for differential recruitment cells from nuclear DNA damage and apoptosis. PLoS ONE of telomeric proteins. Science 2008; 319: 1092–1096. 2013; 8: e52989.

Cell Discovery | www.nature.com/celldisc Studying telosome subunits using inducible knockout cell lines 16

37 Chen LY, Zhang Y, Zhang Q et al. Mitochondrial locali- 56 Sternberg SH, Richter H, Charpentier E, Qimron U. zation of telomeric protein TIN2 links telomere regulation Adaptation in CRISPR-Cas systems. Mol Cell 2016; 61: to metabolic control. Mol Cell 2012; 47: 839–850. 797–808. 38 Sarek G, Marzec P, Margalef P, Boulton SJ. Molecular 57 Amitai G, Sorek R. CRISPR-Cas adaptation: insights into basis of telomere dysfunction in human genetic diseases. the mechanism of action. Nat Rev Microbiol 2016; 14: Nat Struct Mol Biol 2015; 22: 867–874. 67–76. 39 Cong L, Ran FA, Cox D et al. Multiplex genome engi- 58 Takai H, Smogorzewska A, de Lange T. DNA damage foci neering using CRISPR/Cas systems. Science 2013; 339: at dysfunctional telomeres. Curr Biol 2003; 13: 1549–1556. 819–823. 59 Karlseder J, Broccoli D, Dai Y, Hardy S, de Lange T. 40 Mali P, Yang L, Esvelt KM et al. RNA-guided - and ATM-dependent apoptosis induced by telomeres human genome engineering via Cas9. Science 2013; 339: lacking TRF2. Science 1999; 283: 1321–1325. 823–826. 60 Guo X, Deng Y, Lin Y et al. Dysfunctional telomeres 41 Wiedenheft B, Sternberg SH, Doudna JA. RNA-guided activate an ATM-ATR-dependent DNA damage genetic silencing systems in bacteria and archaea. Nature response to suppress tumorigenesis. EMBO J 2007; 26: 2012; 482: 331–338. 4709–4719. 42 Sander JD, Joung JK. CRISPR-Cas systems for editing, 61 Denchi EL, de Lange T. Protection of telomeres through regulating and targeting genomes. Nat Biotechnol 2014; 32: independent control of ATM and ATR by TRF2 347–355. and POT1. Nature 2007; 448: 1068–1071. 43 Zhang F, Wen Y, Guo X. CRISPR/Cas9 for genome 62 Wu L, Multani AS, He H et al. Pot1 deficiency initiates editing: progress, implications and challenges. Hum Mol DNA damage checkpoint activation and aberrant homo- Genet 2014; 23:40–46. logous recombination at telomeres. Cell 2006; 126: 44 Qi LS, Larson MH, Gilbert LA et al. Repurposing 49–62. CRISPR as an RNA-guided platform for sequence-specific 63 Hockemeyer D, Daniels JP, Takai H, de Lange T. Recent control of gene expression. Cell 2013; 152: 1173–1183. expansion of the telomeric complex in rodents: two distinct 45 Jinek M, East A, Cheng A et al. RNA-programmed gen- POT1 proteins protect mouse telomeres. Cell 2006; 126: ome editing in human cells. Elife 2013; 2: e00471. 63–77. 46 Jinek M, Chylinski K, Fonfara I et al. A programmable 64 He H, Multani AS, Cosme-Blanco W et al. POT1b protects dual-RNA-guided DNA endonuclease in adaptive bacterial telomeres from end-to-end chromosomal fusions and immunity. Science 2012; 337: 816–821. aberrant homologous recombination. EMBO J 2006; 25: 47 Haurwitz RE, Jinek M, Wiedenheft B, Zhou K, Doudna JA. 5180–5190. Sequence- and structure-specific RNA processing by a 65 Palm W, Hockemeyer D, Kibe T, de Lange T. Functional CRISPR endonuclease. Science 2010; 329: 1355–1358. dissection of human and mouse POT1 proteins. Mol Cell 48 Chiang YJ, Kim SH, Tessarollo L, Campisi J, Hodes RJ. Biol 2009; 29: 471–482. Telomere-associated protein TIN2 is essential for early 66 Hockemeyer D, Palm W, Else T et al. Telomere protection embryonic development through a telomerase-independent by mammalian Pot1 requires interaction with Tpp1. Nat pathway. Mol Cell Biol 2004; 24: 6631–6634. Struct Mol Biol 2007; 14: 754–761. 49 Karlseder J, Kachatrian L, Takai H et al. Targeted deletion 67 Churikov D, Wei C, Price CM. Vertebrate POT1 reveals an essential function for the telomere length restricts G-overhang length and prevents activation of a regulator Trf1. Mol Cell Biol 2003; 23: 6533–6541. telomeric DNA damage checkpoint but is dispensable 50 Celli GB, de Lange T. DNA processing is not required for for overhang protection. Mol Cell Biol 2006; 26: ATM-mediated telomere damage response after TRF2 6971–6982. deletion. Nat Cell Biol 2005; 7: 712–718. 68 Baumann P, Price C. Pot1 and telomere maintenance. 51 Orban PC, Chui D, Marth JD. Tissue- and site-specific FEBS Lett 2010; 584: 3779–3784. DNA recombination in transgenic mice. Proc Natl Acad Sci 69 Sexton AN, Regalado SG, Lai CS et al. Genetic USA 1992; 89: 6861–6865. and molecular identification of three human TPP1 func- 52 Sauer B. Inducible gene targeting in mice using the Cre/ tions in telomerase action: recruitment, activation, and lox system. Methods 1998; 14: 381–392. homeostasis set point regulation. Genes Dev 2014; 28: 53 Gu H, Zou YR, Rajewsky K. Independent control of 1885–1899. immunoglobulin switch recombination at individual switch 70 Nakashima M, Nandakumar J, Sullivan KD, Espinosa JM, regions evidenced through Cre-loxP-mediated gene target- Cech TR. Inhibition of telomerase recruitment and cancer ing. Cell 1993; 73: 1155–1164. cell death. J Biol Chem 2013; 288: 33171–33180. 54 Wang T, Wei JJ, Sabatini DM, Lander ES. Genetic screens 71 Lue NF, Yu EY, Lei M. A popular engagement at the ends. in human cells using the CRISPR-Cas9 system. Science Nat Struct Mol Bio 2013; 20:10–12. 2014; 343:80–84. 72 de Lange T. Protection of mammalian telomeres. Oncogene 55 Wright AV, Nuñez JK, Doudna JA. Biology and applica- 2002; 21: 532–540. tions of CRISPR systems: harnessing nature’s toolbox for 73 Greider CW. Telomeres do D-loop-T-loop. Cell 1999; 97: genome engineering. Cell 2016; 164:29–44. 419–422.

Cell Discovery | www.nature.com/celldisc Hyeung Kim et al. 17

74 Neidle S, Parkinson GN. The structure of telomeric DNA. 88 Kabir S, Hockemeyer D, de Lange T. TALEN gene Curr Opin Struct Biol 2003; 13: 275–283. knockouts reveal no requirement for the conserved human 75 Flynn RL, Centore RC, O’Sullivan RJ et al. TERRA shelterin protein Rap1 in telomere protection and length and hnRNPA1 orchestrate an RPA-to-POT1 switch on regulation. Cell Rep 2014; 9: 1273–1280. telomeric single-stranded DNA. Nature 2011; 471: 89 Baumann P. Are mouse telomeres going to pot? Cell 2006; 532–536. 126:33–36. 76 Carneiro T, Khair L, Reis CC et al. Telomeres avoid end 90 Shalem O, Sanjana NE, Hartenian E et al. Genome-scale detection by severing the checkpoint signal transduction CRISPR-Cas9 knockout screening in human cells. Science pathway. Nature 2010; 467: 228–232. 2014; 343:84–87. 77 Barrientos KS, Kendellen MF, Freibaum BD et al. Distinct 91 Emanuele MJ, Elia AE, Xu Q et al. Global identification of functions of POT1 at telomeres. Mol Cell Biol 2008; 28: modular cullin-RING ligase substrates. Cell 2011; 147: 5251–5264. 459–474. 78 Takai KK, Kibe T, Donigian JR, Frescas D, de Lange T. 92 Yang D, Xiong Y, Kim H et al. Human telomeric proteins Telomere protection by TPP1/POT1 requires tethering occupy selective interstitial sites. Cell Res 2011; 21: to TIN2. Mol Cell 2011; 44: 647–659. 1013–1027. 79 Kibe T, Osawa GA, Keegan CE, de Lange T. Telomere 93 He Q, Kim H, Huang R et al. The Daxx/Atrx complex protection by TPP1 is mediated by POT1a and POT1b. Mol protects tandem repetitive elements during DNA hypo- Cell Biol 2010; 30: 1059–1066. methylation by promoting H3K9 trimethylation. Cell Stem 80 Tejera AM, Stagno d’Alcontres M, Thanasoula M et al. Cell 2015; 17: 273–286. TPP1 is required for TERT recruitment, telomere elonga- 94 Ouellette MM, Liao M, Herbert BS et al. Subsenescent tion during nuclear reprogramming, and normal skin telomere lengths in fibroblasts immortalized by limiting development in mice. Dev Cell 2010; 18: 775–789. amounts of telomerase. J Biol Chem 2000; 275: 81 He H, Wang Y, Guo X et al. Pot1b deletion and telomerase 10072–10076. haploinsufficiency in mice initiate an ATR-dependent 95 McElligott R, Wellinger RJ. The terminal DNA structure DNA damage response and elicit phenotypes resembling of mammalian chromosomes. EMBO J 1997; 16: dyskeratosis congenita. Mol Cell Biol 2009; 29: 229–240. 3705–3714. 82 Yang Q, Zhang R, Horikawa I et al. Functional diversity of 96 von Rundstedt FC, Rajapakshe K, Ma J et al. Integrative human protection of telomeres 1 isoforms in telomere pathway analysis of metabolic signature in bladder cancer: protection and cellular . Cancer Res 2007; 67: a linkage to The Cancer Genome Atlas Project and Pre- 11677–11686. diction of Survival. J Urol 2016; 195: 1911–1919. 83 Martínez P, Gómez-López G, García F et al. RAP1 pro- tects from obesity through its extratelomeric role regulating (Supplementary Information is linked to the online version of the – gene expression. Cell Rep 2013; 3: 2059 2074. paper on the Discovery website.) 84 Martinez-Outschoorn UE, Peiris-Pagés M, Pestell RG, Sotgia F, Lisanti MP. Cancer metabolism: a therapeutic perspective. Nat Rev Clin Oncol 2016; 14:11–31. This work is licensed under a Creative Commons 85 Boroughs LK, DeBerardinis RJ. Metabolic pathways pro- Attribution 4.0 International License. The images or moting cancer cell survival and growth. Nat Cell Biol 2015; other third party material in this article are included in the article’s 17: 351–359. Creative Commons license, unless indicated otherwise in the 86 Martinez P, Thanasoula M, Carlos AR et al. Mammalian credit line; if the material is not included under the Creative Rap1 controls telomere function and gene expression Commons license, users will need to obtain permission from the through binding to telomeric and extratelomeric sites. Nat license holder to reproduce the material. To view a copy of this Cell Biol 2010; 12: 768–780. license, visit http://creativecommons.org/licenses/by/4.0/ 87 Sfeir A, Kabir S, van Overbeek M, Celli GB, de Lange T. Loss of Rap1 induces telomere recombination in the absence of NHEJ or a DNA damage signal. Science 2010; 327: 1657–1661. © The Author(s) 2017

Cell Discovery | www.nature.com/celldisc